AlphaEvolve (Novikov et al. 2025) | Google DeepMind

English Podcast starts at 00:00:12Bengali Podcast Starts at 00:19:23🎙️ Welcome back to another episode of Revise and Resubmit!Your favorite podcast where footnotes get the spotlight, white papers go mainstream, and the future of science sounds like something straight out of science fiction... but real. 😎📄🧠🧬 Today’s headline? It's not just a paper. It's a plot twist in the history of computation.📢 Introducing: “AlphaEvolve: A Coding Agent for Scientific and Algorithmic Discovery” — hot off the research press from Google DeepMind, published on May 14, 2025.Authored by Alexander Novikov and team, this isn’t just another AI tool. It’s not a chatbot. It’s not a code assistant.🔥 It’s an evolutionary coding agent.An autonomous swarm of LLMs, working in loops, improving themselves, rewriting algorithms, optimizing circuits, and yes — even speeding up the training of their own kind. 🤯Let that sink in for a second.AlphaEvolve didn’t just solve problems.💥 It discovered a new matrix multiplication method — breaking a 56-year-old record since Strassen’s algorithm.🧩 It simplified hardware circuits.📊 It turbocharged Google's data center schedulers.And like a brainy ouroboros, it made the LLM behind AlphaEvolve better at being AlphaEvolve. 🤖✨♻️This white paper doesn’t read like documentation. It reads like the origin story of AI-powered scientific discovery. It’s the future, written in Python and published in LaTeX.But here’s the real kicker…👀 If algorithms can evolve themselves into something better — where does human ingenuity fit in the equation?🎧 If that question makes your neurons fire, hit subscribe on our Spotify channel “Revise and Resubmit”, and don’t forget to check out the visuals, explainers, and behind-the-scenes breakdowns on our YouTube channel “Weekend Researcher”. 🎥🧪We’re also streaming on Amazon Prime and Apple Podcast — because your curiosity deserves a front-row seat, wherever you are. 🚀📱So — pull up a chair.Plug in those headphones.Because AlphaEvolve didn’t just change the code.It rewrote the rules.ReferenceAlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms. (2025, May 14). Google DeepMind. https://deepmind.google/discover/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/Funsearch paper:Romera-Paredes, B., Barekatain, M., Novikov, A. et al. Mathematical discoveries from program search with large language models. Nature 625, 468–475 (2024). https://doi.org/10.1038/s41586-023-06924-6‌Youtube channel link https://www.youtube.com/@weekendresearcherSupport us on Patreonhttps://patreon.com/weekendresearcher